Armington, P.S., “A Theory of Demand for Products Distinguished by Place of Production,” IMF Staff Papers No. 16, (1969) pp. 159–78.
Bertola, G., and L.E.O. Svensson, “Stochastic Devaluation Risk and the Empirical Fit of Target Zone Models,” NBER Working Paper No. 3955, (1991), forthcoming in Review of Economic Studies.
Boughton, J.M., “The Monetary Approach to Exchange Rates: What Now Remains?” Princeton Studies in International Finance No. 171, (1988).
Caramazza, F., “French-German Interest Rate Differentials and Time-Varying Realignment Risk,” International Monetary Fund, Staff Papers No. 40, (1993), pp. 567–583.
Cumby, R.E., J. Huizinga, and M. Obstfeld, “Two-Step Two-Stage Least Squares Estimation in Models with Rational Expectations,” Journal of Econometrics No. 21, (1983), pp. 333–355.
Honohan, P., and C. Conroy, “Sterling Movements and Irish Pound Interest Rates,” Economic and Social Research Institute, mimeo, (1993).
Hodrik, R.J., The Empirical Evidence on the Efficiency of Forward and Futures Foreign Exchange Markets, London, Harwood, (1987).
McGuire, M., “Irish Trade Exposure to Fluctuations in the Irish Pound/Sterling Exchange Rate,” Economic Affairs Department, Central Bank of Ireland, mimeo, (1993).
Newey, W.K., and K. West, “A Simple, Positive Semi-Definite, Heteroskedasticity and Autocorrelation Consistent Covariance Matrix,” Econometrica No. 55, (1987), pp. 703–708.
Rose, A., and L.E.O. Svensson, “Expected and Predicted Realignments: The FF/DM Exchange Rate during the EMS,” Board of Governors of the Federal Reserve, International Finance Discussion Papers, No. 395, (1991).
Svensson, L.E.O., “The Foreign Exchange Risk Premium in a Target Zone with Devaluation Risk,” Journal of International Economics No. 33, (1992), pp. 21–40.
Svensson, L.E.O., “Assessing Target Zone Credibility: Mean Reversion and Devaluation Expectations in the EMS,” European Economic Review, No. 37, (1993), pp. 763–802.
Taylor, M.P., “Exchange Rate Behavior Under Alternative Exchange Rate Arrangements,” paper presented at the Fiftieth Anniversary Conference of the Princeton Essays in International Finance, Princeton University, (1993)
Walsh, B., “Credibility, Interest Rates and the ERM: The Irish Experience, 1986-92,” University College, Dublin, Department of Economics WP 93/1, (1993).
Wickham, P., “A Cautionary Note on the Use of Exchange Rate Indicators,” IMF, Paper on Policy Analysis and Assessment 93/5, (1993).
I thank J. Boughton, B. Chadha, P. Clark, D. Harrison, F. Lakwjik, and E. Prasad for various comments and suggestions. The opinions expressed here do not necessarily reflect the views of the International Monetary Fund.
While the Bertola-Svensson technique has been suggested in the context of a model of an exchange rate band, the same technique is applicable to a variety of seldom-moving targets, including crawling exchange rate bands (such as those implemented in Israel, Mexico and Chile) and gold standard intervention rules.
Data from free-floating exchange rates traditionally provides evidence against UIP (see, for instance, the surveys by Hodrick (1987) and Boughton (1988)). Svensson (1992), however, has shown that risk premia for currencies regulated within a relatively small target zone are likely to be small: even for extreme parameter values and for currencies subject to realignment risk, the premium should be not greater than 1/5 of the whole interest differential. Thus, apparent interest parity failures for ERM currencies are more likely to reflect a special type of ‘peso problem’, i.e., investors’ expectation of large realignments which may or may not occur in any given sample. The technique described in the text is aimed exactly at measuring investors’ expectations of these infrequent events.
More precisely, Et[ΔCi,t+Δt] gives an estimate of the expected size of the realignment multiplied by its probability rate. Further restrictions on the behavior of the exchange rate at the time of a realignment would be necessary to disentangle these two components. For the present analysis, as in much of the related literature, only the product of the two matters.
Most of the transaction costs faced by investors borrowing on the domestic market and lending on the Euro-market can be captured by defining the interest differential as the difference between the offshore bid rate and the onshore ask rate.
Since the regressors in equation (5) are defined over overlapping intervals of 65 observations, the regression errors follow a moving average process of order 65 even if the underlying errors are i.i.d., and therefore exhibit high serial correlation of the residuals. The 2S2SLS estimator corrects for serial correlation and heteroskedasticity by weighing sample covariances at different lags as described in Newey-West (1987). Note that consistent estimation of the standard error of the estimates is necessary to guide selection of the final regression, but is irrelevant for the analysis that follows.
The simple statistical model described above is not designed to describe jumps of the exchange rate position in the band that may occur at the time of a realignment. Therefore, the estimated devaluation series is inclusive of changes of the exchange rate inside the band at the moment of a realignment. See Svensson (1993) for a detailed discussion of this point.
Classical statistical analysis is appropriate for hypothesis testing: the position of the exchange rate in its ERM band is defined over a finite interval, and hence cannot contain a unit root, while interest differentials among industrial countries are well known to be stationary (for Ireland, see Honahan and Conroy (1993)).
The shift dummies may capture other effects than the institutional changes they are designed for. The significance of the lira dummy in the Dutch guilder model, for instance, is more likely to capture the regime shift induced by the ERM crisis in September 1992 than the effects of the withdrawal of the lira from the system.
Note that the level of the estimated expected devaluation series is defined with respect to the specific (three-month) term of the contract, and measures expected devaluation by multiplying a probability of devaluation by the expected size of the devaluation. Illustratively, during the period October 1992-January 1993, investors were expecting a 4-6 percent devaluation of the Irish pound over the next three-months, which may be interpreted as a sure devaluation of 4-6 percent, or as a 50 percent probability attached to a realignment of 8-12 percent and a 50 percent probability attached to no realignment at all, etc.
Recall that sterling was fluctuating in a wide 6 percent band during its participation in the ERM.
The theoretical model underlying the analysis of McGuirk (1987) is the imperfect substitute model of Armington (1969), where the demand for each country’s output is obtained by a two-step maximization with a CES utility function.
These weights mainly reflect the overall degree of openness of the economy, which has remained virtually the same in Ireland (at 115 percent) when measured by the share of exports plus imports over GDP. To the extent that Ireland’s exports have grown slightly more than its imports, the approximation in the text may tend to slightly overestimate the current importance of domestic competition with respect to competition in export markets.
The British Central Statistical Office estimates overall import penetration in the United Kingdom in 1989 at 95 percent in office machinery and data processing, 52 percent in electrical and electronic engineering and 60 percent in instrument engineering. In the chemical sector, British domestic producers tend to be concentrated in the industrial chemical sector, rather than in pharmaceuticals, where much of Ireland’s chemical production is concentrated. See McGuire (1993) for a discussion.
Both series have been normalized by measuring their level as a percentage deviation from the mean level in the period.
Besides the greater timeliness and accuracy of CPIs with respect to other deflators, manufacturing-based deflators (such as common unit labor cost deflators) tend to be particularly unreliable in Ireland. The shift of the Irish manufacturing basis toward more capital intensive high-technology production implies that the large decline of unit labor costs in Ireland since the early 1980s reflects more a compositional effect, rather than the relative decline of unit labor costs of a constant basket of goods.
CPI data is from the International Financial Statistics of the IMF. When unavailable (as in the case of Ireland), monthly series were obtained as weighted moving averages from the published quarterly series.
June 1992 represents a natural benchmark to date the beginning of the Irish pound crisis. That month marks the reversal of the flow of Irish official reserves, the upturn of interest rate differentials, the beginning of sterling’s slide within its ERM band and the consequent beginning of the appreciation of Ireland’s real exchange rate.
Several recent papers have provided strong evidence in favor of a long-run tendency of most exchange rates to revert to Purchasing Power Parity (PPP), based on cointegration methods and long series of historical data. See Taylor (1993) for a survey of the literature.
Note that all exchange rates in (6) and (7) are defined in terms of currency N (the Deutsche mark), so that (x0,t-xi,t) represents the (log) exchange rate of the Irish pound with currency i.
More than 90 percent of Ireland’s trade is accounted for by including in the sample the six ERM currencies studied in Section III, the dollar and the yen (in addition to the reference Deutsche mark).
One reason why measurement errors may not be a problem in the present specification is that these errors are likely to largely wash out when forming the composite indicator of foreign expected devaluation. Note that measurement errors in the dependent variable do not affect consistency of the estimation, since they are absorbed in the regression error.